Prognostic risk stratification of gliomas using deep learning in digital pathology images.

Background: Evaluation of tumor-tissue images stained with hematoxylin and eosin (H&E) is pivotal in diagnosis, yet only a fraction of the rich phenotypic information is considered for clinical care. Here, we propose a survival deep learning (SDL) framework to extract this information to predict glioma survival. Methods: Digitized whole slide images were downloaded from The Cancer Genome Atlas (TCGA) for 766 diffuse glioma patients, including isocitrate dehydrogenase (IDH)-mutant/1p19q-codeleted oligodendroglioma, IDH-mutant/1p19q-intact astrocytoma, and IDH-wildtype astrocytoma/glioblastoma. Our SDL framework employs a residual convolutional neural network with a survival model to predict patient risk from H&E-stained whole-slide images. We used statistical sampling techniques and randomized the transformation of images to address challenges in learning from histology images. The SDL risk score was evaluated in traditional and recursive partitioning (RPA) survival models. Results: The SDL risk score demonstrated substantial univariate prognostic power (median concordance index of 0.79 [se: 0.01]). After adjusting for age and World Health Organization 2016 subtype, the SDL risk score was significantly associated with overall survival (OS; hazard ratio = 2.45; 95% CI: 2.01 to 3.00). Four distinct survival risk groups were characterized by RPA based on SDL risk score, IDH status, and age with markedly different median OS ranging from 1.03 years to 14.14 years. Conclusions: The present study highlights the independent prognostic power of the SDL risk score for objective and accurate prediction of glioma outcomes. Further, we show that the RPA delineation of patient-specific risk scores and clinical prognostic factors can successfully demarcate the OS of glioma patients.

Improving the noninvasive classification of glioma genetic subtype with deep learning and diffusion-weighted imaging.

BACKGROUND: Diagnostic classification of diffuse gliomas now requires an assessment of molecular features, often including IDH-mutation and 1p19q-codeletion status. Because genetic testing requires an invasive process, an alternative noninvasive approach is attractive, particularly if resection is not recommended. The goal of this study was to evaluate the effects of training strategy and incorporation of biologically relevant images on predicting genetic subtypes with deep learning. METHODS: Our dataset consisted of 384 patients with newly diagnosed gliomas who underwent preoperative MRI with standard anatomical and diffusion-weighted imaging, and 147 patients from an external cohort with anatomical imaging. Using tissue samples acquired during surgery, each glioma was classified into IDH-wildtype (IDHwt), IDH-mutant/1p19q-noncodeleted (IDHmut-intact), and IDH-mutant/1p19q-codeleted (IDHmut-codel) subgroups. After optimizing training parameters, top performing convolutional neural network (CNN) classifiers were trained, validated, and tested using combinations of anatomical and diffusion MRI with either a 3-class or tiered structure. Generalization to an external cohort was assessed using anatomical imaging models. RESULTS: The best model used a 3-class CNN containing diffusion-weighted imaging as an input, achieving 85.7% (95% CI: [77.1, 100]) overall test accuracy and correctly classifying 95.2%, 88.9%, 60.0% of the IDHwt, IDHmut-intact, and IDHmut-codel tumors. In general, 3-class models outperformed tiered approaches by 13.5%-17.5%, and models that included diffusion-weighted imaging were 5%-8.8% more accurate than those that used only anatomical imaging. CONCLUSION: Training a classifier to predict both IDH-mutation and 1p19q-codeletion status outperformed a tiered structure that first predicted IDH-mutation, then 1p19q-codeletion. Including apparent diffusion coefficient (ADC), a surrogate marker of cellularity, more accurately captured differences between subgroups.

Interactions of Age and Blood Immune Factors and Noninvasive Prediction of Glioma Survival.

BACKGROUND: Tumor-based classification of human glioma portends patient prognosis, but considerable unexplained survival variability remains. Host factors (eg, age) also strongly influence survival times, partly reflecting a compromised immune system. How blood epigenetic measures of immune characteristics and age augment molecular classifications in glioma survival has not been investigated. We assess the prognostic impact of immune cell fractions and epigenetic age in archived blood across glioma molecular subtypes for the first time. METHODS: We evaluated immune cell fractions and epigenetic age in archived blood from the University of California San Francisco Adult Glioma Study, which included a training set of 197 patients with IDH-wild type, 1p19q intact, TERT wild type (IDH/1p19q/TERT-WT) glioma, an evaluation set of 350 patients with other subtypes of glioma, and 454 patients without glioma. RESULTS: IDH/1p19q/TERT-WT patients had lower lymphocyte fractions (CD4+ T, CD8+ T, natural killer, and B cells) and higher neutrophil fractions than people without glioma. Recursive partitioning analysis delineated 4 statistically significantly different survival groups for patients with IDH/1p19q/TERT-WT based on an interaction between chronological age and 2 blood immune factors, CD4+ T cells, and neutrophils. Median overall survival ranged from 0.76 years (95% confidence interval = 0.55-0.99) for the worst survival group (n = 28) to 9.72 years (95% confidence interval = 6.18 to not available) for the best (n = 33). The recursive partitioning analysis also statistically significantly delineated 4 risk groups in patients with other glioma subtypes. CONCLUSIONS: The delineation of different survival groups in the training and evaluation sets based on an interaction between chronological age and blood immune characteristics suggests that common host immune factors among different glioma types may affect survival. The ability of DNA methylation-based markers of immune status to capture diverse, clinically relevant information may facilitate noninvasive, personalized patient evaluation in the neuro-oncology clinic.

Detection of glioma infiltration at the tumor margin using quantitative stimulated Raman scattering histology.

In the management of diffuse gliomas, the identification and removal of tumor at the infiltrative margin remains a central challenge. Prior work has demonstrated that fluorescence labeling tools and radiographic imaging are useful surgical adjuvants with macroscopic resolution. However, they lose sensitivity at the tumor margin and have limited clinical utility for lower grade histologies. Fiber-laser based stimulated Raman histology (SRH) is an optical imaging technique that provides microscopic tissue characterization of unprocessed tissues. It remains unknown whether SRH of tissues taken from the infiltrative glioma margin will identify microscopic residual disease. Here we acquired glioma margin specimens for SRH, histology, and tumor specific tissue characterization. Generalized linear mixed models were used to evaluate agreement. We find that SRH identified residual tumor in 82 of 167 margin specimens (49%), compared to IHC confirming residual tumor in 72 of 128 samples (56%), and H&E confirming residual tumor in 82 of 169 samples (49%). Intraobserver agreements between all 3 modalities were confirmed. These data demonstrate that SRH detects residual microscopic tumor at the infiltrative glioma margin and may be a promising tool to enhance extent of resection.

Recurrent tumor and treatment-induced effects have different MR signatures in contrast enhancing and non-enhancing lesions of high-grade gliomas.

BACKGROUND: Differentiating treatment-induced injury from recurrent high-grade glioma is an ongoing challenge in neuro-oncology, in part due to lesion heterogeneity. This study aimed to determine whether different MR features were relevant for distinguishing recurrent tumor from the effects of treatment in contrast-enhancing lesions (CEL) and non-enhancing lesions (NEL). METHODS: This prospective study analyzed 291 tissue samples (222 recurrent tumor, 69 treatment-effect) with known coordinates on imaging from 139 patients who underwent preoperative 3T MRI and surgery for a suspected recurrence. 8 MR parameter values were tested from perfusion-weighted, diffusion-weighted, and MR spectroscopic imaging at each tissue sample location for association with histopathological outcome using generalized estimating equation models for CEL and NEL tissue samples. Individual cutoff values were evaluated using receiver operating characteristic curve analysis with 5-fold cross-validation. RESULTS: In tissue samples obtained from CEL, elevated relative cerebral blood volume (rCBV) was associated with the presence of recurrent tumor pathology (P < 0.03), while increases in normalized choline (nCho) and choline-to-NAA index (CNI) were associated with the presence of recurrent tumor pathology in NEL tissue samples (P < 0.008). A mean CNI cutoff value of 2.7 had the highest performance, resulting in mean sensitivity and specificity of 0.61 and 0.81 for distinguishing treatment-effect from recurrent tumor within the NEL. CONCLUSION: Although our results support prior work that underscores the utility of rCBV in distinguishing the effects of treatment from recurrent tumor within the contrast enhancing lesion, we found that metabolic parameters may be better at differentiating recurrent tumor from treatment-related changes in the NEL of high-grade gliomas.

Association of Maximal Extent of Resection of Contrast-Enhanced and Non-Contrast-Enhanced Tumor With Survival Within Molecular Subgroups of Patients With Newly Diagnosed Glioblastoma.

Importance: Per the World Health Organization 2016 integrative classification, newly diagnosed glioblastomas are separated into isocitrate dehydrogenase gene 1 or 2 (IDH)-wild-type and IDH-mutant subtypes, with median patient survival of 1.2 and 3.6 years, respectively. Although maximal resection of contrast-enhanced (CE) tumor is associated with longer survival, the prognostic importance of maximal resection within molecular subgroups and the potential importance of resection of non-contrast-enhanced (NCE) disease is poorly understood. Objective: To assess the association of resection of CE and NCE tumors in conjunction with molecular and clinical information to develop a new road map for cytoreductive surgery. Design, Setting, and Participants: This retrospective, multicenter cohort study included a development cohort from the University of California, San Francisco (761 patients diagnosed from January 1, 1997, through December 31, 2017, with 9.6 years of follow-up) and validation cohorts from the Mayo Clinic (107 patients diagnosed from January 1, 2004, through December 31, 2014, with 5.7 years of follow-up) and the Ohio Brain Tumor Study (99 patients with data collected from January 1, 2008, through December 31, 2011, with a median follow-up of 10.9 months). Image accessors were blinded to patient groupings. Eligible patients underwent surgical resection for newly diagnosed glioblastoma and had available survival, molecular, and clinical data and preoperative and postoperative magnetic resonance images. Data were analyzed from November 15, 2018, to March 15, 2019. Main Outcomes and Measures: Overall survival. Results: Among the 761 patients included in the development cohort (468 [61.5%] men; median age, 60 [interquartile range, 51.6-67.7] years), younger patients with IDH-wild-type tumors and aggressive resection of CE and NCE tumors had survival similar to that of patients with IDH-mutant tumors (median overall survival [OS], 37.3 [95% CI, 31.6-70.7] months). Younger patients with IDH-wild-type tumors and reduction of CE tumor but residual NCE tumors fared worse (median OS, 16.5 [95% CI, 14.7-18.3] months). Older patients with IDH-wild-type tumors benefited from reduction of CE tumor (median OS, 12.4 [95% CI, 11.4-14.0] months). The results were validated in the 2 external cohorts. The association between aggressive CE and NCE in patients with IDH-wild-type tumors was not attenuated by the methylation status of the promoter region of the DNA repair enzyme O6-methylguanine-DNA methyltransferase. Conclusions and Relevance: This study confirms an association between maximal resection of CE tumor and OS in patients with glioblastoma across all subgroups. In addition, maximal resection of NCE tumor was associated with longer OS in younger patients, regardless of IDH status, and among patients with IDH-wild-type glioblastoma regardless of the methylation status of the promoter region of the DNA repair enzyme O6-methylguanine-DNA methyltransferase. These conclusions may help reassess surgical strategies for individual patients with newly diagnosed glioblastoma.

Safety of Outpatient Anterior Cervical Discectomy and Fusion: A Systematic Review and Meta-Analysis.

BACKGROUND: Anterior cervical discectomy and fusion (ACDF) is being increasingly offered on an outpatient basis. However, the safety profile of outpatient ACDF remains poorly defined. OBJECTIVE: To review the medical literature on the safety of outpatient ACDF. METHODS: We systematically reviewed the literature for articles published before April 1, 2018, describing outpatient ACDF and associated complications, including incidence of reoperation, stroke, thrombolytic events, dysphagia, hematoma, and mortality. A random-effects analysis was performed comparing complications between the inpatient and outpatient groups. RESULTS: We identified 21 articles that satisfied the selection criteria, of which 15 were comparative studies. Most of the existing studies were retrospective, with a lack of level I or II studies on this topic. We found no statistically significant difference between inpatient and outpatient ACDF in overall complications, incidence of stroke, thrombolytic events, dysphagia, and hematoma development. However, patients undergoing outpatient ACDF had lower reported reoperation rates (P < .001), mortality (P < .001), and hospitalization duration (P < .001). CONCLUSION: Our meta-analysis indicates that there is a lack of high level of evidence studies regarding the safety of outpatient ACDF. However, the existing literature suggests that outpatient ACDF can be safe, with low complication rates comparable to inpatient ACDF in well-selected patients. Patients with advanced age and comorbidities such as obesity and significant myelopathy are likely not suitable for outpatient ACDF. Spine surgeons must carefully evaluate each patient to decide whether outpatient ACDF is a safe option. Higher quality, large prospective randomized control trials are needed to accurately demonstrate the safety profile of outpatient ACDF.

An independently validated nomogram for isocitrate dehydrogenase-wild-type glioblastoma patient survival.

BACKGROUND: In 2016, the World Health Organization reclassified the definition of glioblastoma (GBM), dividing these tumors into isocitrate dehydrogenase (IDH)-wild-type and IDH-mutant GBM, where the vast majority of GBMs are IDH-wild-type. Nomograms are useful tools for individualized estimation of survival. This study aimed to develop and independently validate a nomogram for IDH-wild-type patients with newly diagnosed GBM. METHODS: Data were obtained from newly diagnosed GBM patients from the Ohio Brain Tumor Study (OBTS) and the University of California San Francisco (UCSF) for diagnosis years 2007-2017 with the following variables: age at diagnosis, sex, extent of resection, concurrent radiation/temozolomide (TMZ) status, Karnofsky Performance Status (KPS), O6-methylguanine-DNA methyltransferase (MGMT) methylation status, and IDH mutation status. Survival was assessed using Cox proportional hazards regression, random survival forests, and recursive partitioning analysis, with adjustment for known prognostic factors. The models were developed using the OBTS data and independently validated using the UCSF data. Models were internally validated using 10-fold cross-validation and externally validated by plotting calibration curves. RESULTS: A final nomogram was validated for IDH-wild-type newly diagnosed GBM. Factors that increased the probability of survival included younger age at diagnosis, female sex, having gross total resection, having concurrent radiation/TMZ, having a high KPS, and having MGMT methylation. CONCLUSIONS: A nomogram that calculates individualized survival probabilities for IDH-wild-type patients with newly diagnosed GBM could be useful to physicians for counseling patients regarding treatment decisions and optimizing therapeutic approaches. Free software for implementing this nomogram is provided: https://gcioffi.shinyapps.io/Nomogram_For_IDH_Wildtype_GBM_H_Gittleman/.

A historical cohort of temporal lobe surgery for medically refractory epilepsy: a systematic review and meta-analysis to guide future nonrandomized controlled trial studies.

OBJECTIVE: Recent trials for temporal lobe epilepsy (TLE) highlight the challenges of investigating surgical outcomes using randomized controlled trials (RCTs). Although several reviews have examined seizure-freedom outcomes from existing data, there is a need for an overall seizure-freedom rate estimated from level I data as investigators consider other methods besides RCTs to study outcomes related to new surgical interventions. METHODS: The authors performed a systematic review and meta-analysis of the 3 RCTs of TLE in adults and report an overall surgical seizure-freedom rate (Engel class I) composed of level I data. An overall seizure-freedom rate was also collected from level II data (prospective cohort studies) for validation. Eligible studies were identified by filtering a published Cochrane meta-analysis of epilepsy surgery for RCTs and prospective studies, and supplemented by searching indexed terms in MEDLINE (January 1, 2012-April 1, 2018). Retrospective studies were excluded to minimize heterogeneity in patient selection and reporting bias. Data extraction was independently reverified and pooled using a fixed-effects model. The primary outcome was overall seizure freedom following surgery. The historical benchmark was applied in a noninferiority study design to compare its power to a single-study cohort. RESULTS: The overall rate of seizure freedom from level I data was 72.4% (55/76 patients, 3 RCTs), which was nearly identical to the overall seizure-freedom rate of 71.7% (1325/1849 patients, 18 studies) from prospective cohorts (z = 0.134, p = 0.89; z-test). Seizure-freedom rates from level I and II studies were consistent over the years of publication (R2 < 0.01, p = 0.73). Surgery resulted in markedly improved seizure-free outcomes compared to medical management (RR 10.82, 95% CI 3.93-29.84, p < 0.01; 2 RCTs). Noninferiority study designs in which the historical benchmark was used had significantly higher power at all difference margins compared to using a single cohort alone (p < 0.001, Bonferroni's multiple comparison test). CONCLUSIONS: The overall rate of seizure freedom for temporal lobe surgery is approximately 70% for medically refractory epilepsy. The small sample size of the RCT cohort underscores the need to move beyond standard RCTs for epilepsy surgery. This historical seizure-freedom rate may serve as a useful benchmark to guide future study designs for new surgical treatments for refractory TLE.

Gait symmetric adaptation: Comparing effects of implicit visual distortion versus split-belt treadmill on aftereffects of adapted step length symmetry.

Understanding gait adaptation is essential for rehabilitation, and visual feedback can be used during gait rehabilitation to develop effective gait training. We have previously shown that subjects can adapt spatial aspects of walking to an implicitly imposed distortion of visual feedback of step length. To further investigate the storage benefit of an implicit process engaged in visual feedback distortion, we compared the robustness of aftereffects acquired by visual feedback distortion, versus split-belt treadmill walking. For the visual distortion trial, we implicitly distorted the visual representation of subjects' gait symmetry, whereas for the split-belt trial, the speed ratio of the two belts was gradually adjusted without visual feedback. After adaptation, the visual feedback or the split-belt perturbation was removed while subjects continued walking, and aftereffects of preserved asymmetric pattern were assessed. We found that subjects trained with visual distortion trial retained aftereffects longest. In response to the larger speed ratio of split-belt walking, the subjects showed an increase in the size of aftereffects compared to the smaller speed ratio, but it steeply decreased over time in all the speed ratios tested. In contrast, the visual distortion group showed much slower decreasing rate of aftereffects, which was evidence of longer storage of an adapted gait pattern. Visual distortion adaptation may involve the interaction and integration of the change in motor strategy and implicit process in sensorimotor adaptation. Although it should be clarified more clearly through further studies, the findings of this study suggest that gait control employs distinct adaptive processes during the visual distortion and split-belt walking and also the level of reliance of an implicit process may be greater in the visual distortion adaptation than the split-belt walking adaptation.

Walking Speed Influences the Effects of Implicit Visual Feedback Distortion on Modulation of Gait Symmetry.

The use of visual feedback in gait rehabilitation has been suggested to promote recovery of locomotor function by incorporating interactive visual components. Our prior work demonstrated that visual feedback distortion of changes in step length symmetry entails an implicit or unconscious adaptive process in the subjects' spatial gait patterns. We investigated whether the effect of the implicit visual feedback distortion would persist at three different walking speeds (slow, self-preferred and fast speeds) and how different walking speeds would affect the amount of adaption. In the visual feedback distortion paradigm, visual vertical bars portraying subjects' step lengths were distorted so that subjects perceived their step lengths to be asymmetric during testing. Measuring the adjustments in step length during the experiment showed that healthy subjects made spontaneous modulations away from actual symmetry in response to the implicit visual distortion, no matter the walking speed. In all walking scenarios, the effects of implicit distortion became more significant at higher distortion levels. In addition, the amount of adaptation induced by the visual distortion was significantly greater during walking at preferred or slow speed than at the fast speed. These findings indicate that although a link exists between supraspinal function through visual system and human locomotion, sensory feedback control for locomotion is speed-dependent. Ultimately, our results support the concept that implicit visual feedback can act as a dominant form of feedback in gait modulation, regardless of speed.